A 3D Geological MPS Modeling Method Based on High-Precision Attitude Simulation

Abstract

This application belongs to the field of geological modeling technology, specifically disclosing a three-dimensional geological MPS modeling method based on high-precision attitude simulation. The method includes: extracting the main extension direction of rock strata from borehole data and establishing a modeling space, discretizing it into grid cells; dividing the modeling space into multiple sub-regions, identifying high-density areas, using a machine learning classifier to predict the stratigraphic properties of unfilled grid cells, generating simulated exploration profiles and constructing training images; establishing borehole spatial relationships through constrained Delaunay triangulation and calculating the prior probability distribution of grid cells to construct a prior model; optimizing the simulation path through a quantitative evaluation system and defining the local scanning range of attitude constraints based on stratigraphic bounding box parameters; and performing multi-point geostatistical MPS simulation according to the optimized simulation path and local scanning range to generate a three-dimensional geological model. This application can construct high-precision three-dimensional geological models based on sparse borehole data.

Description

Technical Field

[0001] This application belongs to the field of geological modeling technology, and more specifically, relates to a three-dimensional geological MPS modeling method based on high-precision attitude simulation. Background Technology

[0002] Three-dimensional geological models serve as a bridge between geological theory and exploration practice, playing an irreplaceable role in the prediction of deep and concealed ore bodies. Traditional two-dimensional maps struggle to effectively represent the complex spatial coupling relationships of ore-controlling elements such as strata, structures, and rock masses. Three-dimensional geological modeling technology, however, integrates heterogeneous data from multiple sources, including geological, geophysical, geochemical, and borehole data, to construct a digital and visual representation of the geological structure of the exploration area. Through three-dimensional spatial analysis, the morphology, occurrence, and spatial distribution of ore bodies can be precisely determined, the structural ore-controlling laws can be quantitatively analyzed, and the dominant paths of ore-forming fluid migration and favorable ore-hosting spaces can be visualized. This not only deepens the understanding of metallogenic systems but also enables precise delineation of prospecting target areas and optimization of prospecting engineering layout, thereby significantly reducing exploration risks and improving prospecting efficiency. It is a core technology for achieving "transparent" exploration and quantitative prediction of mineral resources.

[0003] The sparsity, heterogeneity, and ambiguity of geological data are inherent characteristics of geological exploration and modeling processes. These characteristics directly lead to significant uncertainties in 3D geological models constructed using deterministic interpolation methods. Such models often struggle to objectively depict the complex variations within the internal structures of actual geological bodies, thus limiting their reliable application in areas such as mineral resource prediction and geological hazard assessment. Against this backdrop, Multiple Point Statistics (MPS), as a typical uncertainty modeling method, faces two core challenges in its application to geological modeling: acquiring training images and accurately representing complex geological bodies. Furthermore, MPS methods require a high quantity and diversity of borehole data to adequately reflect the non-stationary characteristics of the study area. However, when dealing with geological structures such as inversions, missing data, duplicates, and thin strata, the fragmentation and spatial heterogeneity of existing borehole data severely restrict the accurate construction of models, further exposing the limitations of MPS methods in applying them to complex geological conditions.

[0004] Therefore, how to construct a high-precision three-dimensional geological model based on sparse borehole data is an urgent problem to be solved. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this application is to provide a three-dimensional geological MPS modeling method based on high-precision attitude simulation, aiming to solve the problem of constructing high-precision three-dimensional geological models based on sparse borehole data.

[0006] To achieve the above objectives, in a first aspect, this application provides a three-dimensional geological MPS modeling method based on high-precision attitude simulation, comprising the following steps: S10, Obtain and extract the main extension direction of the rock strata from the borehole data of the study area, establish a modeling space based on the main extension direction of the rock strata, and then discretize the modeling space into grid cells; S20, based on the main extension direction of the rock strata, the modeling space is divided into multiple sub-regions, high-density regions with more than a preset threshold of boreholes are identified, a machine learning classifier is used in the high-density regions to predict the stratigraphic properties of unfilled grid cells, a simulated exploration profile is generated, and a training image is constructed based on the simulated exploration profile. S30, Based on the borehole data, establish the borehole spatial relationship through constrained Delaunay triangulation, and calculate the prior probability distribution of the mesh cells to construct a prior model; S40, based on the prior model and training images, the simulation path is optimized through a quantitative evaluation system, and the local scanning range of the attitude constraint is defined based on the stratigraphic bounding box parameters; S50 performs multi-point geostatistical (MPS) simulations according to the optimized simulation path and the local scan range constrained by attitude, generating a three-dimensional geological model.

[0007] This application provides a 3D geological MPS modeling method based on high-precision attitude simulation, which has the following advantages: By extracting the main extension direction of rock strata from sparse borehole data and establishing a modeling space, a directional framework is provided for subsequent processing, overcoming the problem of directionality loss caused by data sparsity in traditional methods; by dividing into sub-regions and identifying high-density areas, and combining machine learning to predict the attributes of unfilled units, simulated exploration profiles and training images are automatically generated, reducing the dependence on a large number of prior training images; by constrained triangulation and prior probability calculation, the uncertainty of borehole spatial relationships and stratigraphic attributes is quantified, enhancing the reliability of the prior model; by optimizing the simulation path and local scanning of attitude constraints, high-determinism areas are prioritized and stratigraphic attitude changes are captured, improving simulation efficiency and structural consistency; finally, by integrating these improved features, a high-precision 3D geological model can be constructed based on sparse borehole data when performing multi-point geostatistical simulations, effectively handling the challenges brought by data sparsity and heterogeneity, and improving the accuracy and geological rationality of the modeling results.

[0008] As a further preferred option, step S10 specifically includes: The wellhead coordinates, inclination data, and formation stratification data of all boreholes in the study area were obtained as input data. Principal component analysis is used to extract the main extension direction of rock strata from the spatial distribution of boreholes, and this direction is defined as the X-axis of the modeling space. The modeling boundary is determined based on the principles of two-dimensional geostatistics and bounding box theory. The topographic surface of the study area is fitted using the borehole wellhead coordinates. The modeling boundary is then cut through this surface to establish an initial three-dimensional stratigraphic model containing borehole information. The initial three-dimensional formation model is discretized into grid cells using a regular grid. The size of the grid cell is set according to the borehole density, and each grid cell contains at most one borehole.

[0009] As a further preferred option, step S20 specifically includes: The modeling space is divided into multiple equally spaced sub-regions along the X-axis, and the number of boreholes in each sub-region is counted. Set a threshold for the number of boreholes, and identify sub-regions exceeding this threshold as high-density regions; In high-density areas, a support vector machine classifier is used to predict the stratigraphic properties of unfilled grid cells to obtain complete simulated exploration profile data. The simulated exploration profile data is transformed into a three-dimensional deterministic geological model using a contour reconstruction algorithm, and this deterministic geological model is used as the training image.

[0010] As a further preferred option, step S30 specifically involves: Based on the borehole data, the constrained Delaunay triangulation method is used to establish the spatial relationship of the boreholes, project the boreholes onto a two-dimensional plane and identify high-density areas. For each simulated grid cell, its prior probability distribution is determined by spatial correlation calculation; The influence weight of a single borehole is obtained by integrating the vertical segment of the borehole formation. The influence weights of all boreholes within the triangular region are accumulated, and the prior probability distribution of each unit is obtained through normalization. The formation corresponding to the maximum prior attribute of each unit is taken as its prior formation attribute, thus completing the construction of the prior model.

[0011] As a further preferred embodiment, in S40, the optimized simulation path is specifically as follows: Based on the constrained Delaunay triangulation results, triangles are classified into four categories according to the number of deterministic vertices they contain: fully constrained, double-constrained, single-constrained, and unconstrained. For similar triangles, a comprehensive priority index PI is constructed; The simulation priority is determined by sorting the triangle types and PI values ​​in descending order, ensuring that high-deterministic regions are simulated first.

[0012] As a further preferred embodiment, the formula for calculating the comprehensive priority index PI is as follows: PI = α × ESI + (1 - α) × (1 - ANI) Where ESI is the equilateral similarity index, ANI is the area normalization index, and α is the weighting coefficient; The formula for calculating the Equilateral Similarity Index (ESI) is as follows: ) in, To normalize the interior angle deviation, This refers to the deviation in side length. The formula for calculating the Area Standardized Index (ANI) is as follows:

[0013] in, Let the area of ​​the current triangle be... The minimum area among all triangles, It represents the maximum area among all triangles.

[0014] As a further preferred embodiment, the normalized interior angle deviation... The calculation formula is:

[0015] in, The formula for calculating the interior angle deviation is:

[0016] Among them, A, B, and C are the three interior angles of the triangle; The side length deviation The calculation formula is:

[0017] Where a, b, and c are the lengths of the three sides of the triangle; For the average side length, .

[0018] As a further preferred embodiment, in step S40, the local scanning range defined by the stratigraphic bounding box parameters for attitude constraints includes: for the unit to be simulated, the local scanning area is determined by the bounding box of its stratum and the scaling factor vector; during the scanning process, in addition to the current bounding box, several adjacent bounding boxes are also included to ensure that the sampling mode can fully reflect the stratigraphic variation characteristics.

[0019] In a second aspect, this application provides an electronic device including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method as described in any one of the above descriptions.

[0020] Thirdly, this application provides a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the method as described in any one of the above statements.

[0021] It is understood that the beneficial effects of the second and third aspects mentioned above can be found in the relevant descriptions in the first aspect above, and will not be repeated here. Attached Figure Description

[0022] Figure 1 This is a flowchart of a three-dimensional geological MPS modeling method based on high-precision attitude simulation provided in the embodiments of this application; Figure 2 This is a technical roadmap of the method provided in the embodiments of this application; Figure 3 This is a three-dimensional model of the borehole in the study area provided in the embodiments of this application; Figure 4 This is a training image generation method provided in the embodiments of this application; wherein, (a) is a three-dimensional prior model generated based on the original borehole, (b) is the meshing of the three-dimensional prior model, (c) is an arbitrary high-density region, (d) is the extraction and meshing of the high-density region, (e) is the three-dimensional prior model of the high-density region, (f) is the three-dimensional prior model of the high-density region and the borehole, (g) is the region used for exploration profile simulation, (h) is the simulated exploration profile obtained based on the borehole, and (i) is the generated three-dimensional training image.

[0023] Figure 5 This is a borehole clustering analysis method provided in the embodiments of this application; wherein, (a) is the original distribution of boreholes, (b) is the meshing of the modeling region, (c) is the extraction of high-density borehole regions, (d) is the extraction of borehole information for simulating the profile, (e) is the unconstrained borehole point triangulation result, (f) is the high-density region borehole constraint, (g) is the high-density region triangulation, (h) is the high-density region triangulation constraint, and (i) is the constrained borehole point triangulation result; Figure 6 This application provides a priori attribute calculation method based on autocorrelation function and borehole clustering analysis. Figure 7 This application provides a high-precision attitude extraction method; wherein, (a) is a high-density area borehole, (b) is a three-dimensional training image, (c) is a three-dimensional model of each stratum, (d) is a stratum attitude extraction method based on accurate attitude simulation and combined bounding box, and (e) is a stratum attitude extraction method based on borehole resampling and global bounding box. Figure 8This application provides a local scanning method based on attitude simulation; wherein, (a) is a three-dimensional stratigraphic model and attitude simulation, (b) is a simulated attitude region and a local scanning region, (c) is the scanning ratio of each direction of the scanning region, (d) is a training image of the location of the unit to be simulated, (e) is a simulated attitude region, and (f) is a local scanning region. Figure 9 This is a partial cross-section of the training image provided in the embodiments of this application; Figure 10 This is a three-dimensional stratigraphic model of the study area generated based on this method, provided in the embodiments of this application. Detailed Implementation

[0024] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0025] It should be understood that, in the description of this application, the term "multiple" means two or more, unless otherwise expressly and specifically defined.

[0026] like Figure 1 As shown, this application provides a three-dimensional geological MPS modeling method based on high-precision attitude simulation, including steps S10 to S50, which are detailed below: Step S10 involves acquiring and extracting the main extension direction of the rock strata from the borehole data of the study area, establishing a modeling space based on the main extension direction of the rock strata, and then discretizing the modeling space into grid cells. This step determines the dominant orientation of the geological structure, provides a spatial framework for 3D modeling, and converts the continuous space into discrete cells through gridding, facilitating subsequent data management and calculation.

[0027] Step S20 involves dividing the modeling space into multiple sub-regions based on the main extension direction of the rock strata, identifying high-density areas where the number of boreholes exceeds a preset threshold, using a machine learning classifier in these high-density areas to predict the stratigraphic properties of unfilled grid cells, generating simulated exploration profiles, and constructing training images based on these simulated exploration profiles. This step can focus on data-rich areas, using machine learning to supplement missing sparse data, generating continuous geological profiles, and converting them into training images, providing a structural pattern library for multi-point statistical simulations.

[0028] Step S30: Based on borehole data, spatial relationships between boreholes are established through constrained Delaunay triangulation, and the prior probability distribution of grid cells is calculated to construct a prior model. This step quantifies the spatial correlation between borehole points, derives the probability distribution of formation properties for each grid cell, and injects prior geological knowledge into the simulation process.

[0029] Step S40 involves optimizing the simulation path based on the prior model and training images using a quantitative evaluation system, and defining the local scan range for attitude constraints based on stratigraphic bounding box parameters. This step determines the priority order of simulation units, ensures that highly deterministic regions are processed first, and adapts the spatial variability of stratigraphic attitudes to the local scan range.

[0030] Step S50: Following the optimized simulation path and the local scanning range constrained by attitude, perform multi-point geostatistical (MPS) simulation to generate a three-dimensional geological model. This step integrates the aforementioned constraints and optimization strategies, generating a three-dimensional model that conforms to geological laws through multi-point statistical simulation, thus completing the reconstruction from sparse data to a high-precision structure.

[0031] This application provides a 3D geological MPS modeling method based on high-precision attitude simulation, which has the following advantages: By extracting the main extension direction of rock strata from sparse borehole data and establishing a modeling space, a directional framework is provided for subsequent processing, overcoming the problem of directionality loss caused by data sparsity in traditional methods; by dividing into sub-regions and identifying high-density areas, and combining machine learning to predict the attributes of unfilled units, simulated exploration profiles and training images are automatically generated, reducing the dependence on a large number of prior training images; by constrained triangulation and prior probability calculation, the uncertainty of borehole spatial relationships and stratigraphic attributes is quantified, enhancing the reliability of the prior model; by optimizing the simulation path and local scanning of attitude constraints, high-determinism areas are prioritized and stratigraphic attitude changes are captured, improving simulation efficiency and structural consistency; finally, by integrating these improved features, a high-precision 3D geological model can be constructed based on sparse borehole data when performing multi-point geostatistical simulations, effectively handling the challenges brought by data sparsity and heterogeneity, and improving the accuracy and geological rationality of the modeling results.

[0032] In one embodiment, the technical solution to achieve the above objective can be as follows: In order to solve the problem of constructing a high-precision three-dimensional geological model based on sparse borehole data, this embodiment provides a three-dimensional geological MPS modeling method based on high-precision attitude simulation. This method significantly improves the modeling capability of multi-point geostatistics under sparse data conditions through rock stratum attitude constraints and adaptive optimization strategies.

[0033] Specifically, the technical solution provided in this embodiment is as follows: First, an automatic simulation method for strata attitude that integrates principal component analysis, spatial clustering, and machine learning is proposed to reconstruct geologically significant simulated exploration profiles from discrete borehole data; second, based on these profiles, a contour reconstruction algorithm is used to construct training images; then, a simulation path optimization strategy based on triangulation and a comprehensive priority index is established to achieve priority simulation of high-determinism areas; finally, based on local attitude scanning of stratigraphic bounding boxes, geological structural features are integrated into the multi-point statistical simulation process. The detailed steps are as follows: Step one, the simulation of rock strata attitude, specifically involves: First, principal component analysis (PCA) was used to extract the main extension direction of the rock strata from the discrete borehole data, and this direction was set as the X-axis of the modeling space. The modeling boundary was determined based on two-dimensional geostatistics and bounding box theory. The terrain surface was fitted using the borehole head coordinates, and this surface was used to cut the modeling boundary, establishing an initial three-dimensional stratigraphic model containing borehole information. Subsequently, a regular mesh was used to discretize the model, ensuring that each mesh cell contained at most one borehole.

[0034] The modeling space is divided into equally spaced sub-regions (X=K, K∈[1,10]) along the X-axis. The number of boreholes in each sub-region is counted, and high-density areas are identified based on a preset threshold. In the high-density areas, a machine learning classifier is used to predict the stratigraphic properties of unfilled grid cells, generating complete simulated exploration profiles. Based on these profiles, a contour reconstruction algorithm is used to construct a three-dimensional geological model, which is then used as training images.

[0035] For attitude characterization, a combined bounding box method is adopted for optimization: by constructing a series of bounding boxes between adjacent simulated profiles, and combining them with an interpolation algorithm, an accurate description of the stratigraphic attitude is achieved. Each bounding box is defined by a principal direction vector, a length parameter, and center coordinates. By increasing the number of bounding boxes and introducing an interpolation strategy, the method effectively adapts to the spatial variations in stratigraphic thickness and morphology.

[0036] Finally, based on the gridded model, the MC cube algorithm was used to extract the interfaces of each stratum, completing the three-dimensional quantitative characterization of the strata's attitude. This process establishes a complete technical path from discrete borehole data to a three-dimensional attitude model, providing structural constraints for subsequent MPS simulations.

[0037] Step two, the construction of training images, is as follows: The construction of training images is based on two parts: deterministic constraint data and indirect inference data. First, a deterministic 3D geological model of the rock strata is generated based on the constructed simulated exploration profile. This model reflects the morphological characteristics and layered distribution patterns of the main strata in the study area. Subsequently, this deterministic model is incorporated into the training images as a model sample library, providing structural feature data for subsequent multi-point statistical simulations.

[0038] Step 3, the prior model construction is as follows: The prior model is constructed through the following technical process: First, the borehole data is resampled, and meshing is performed within the established modeling area. The mesh cells corresponding to the borehole locations are directly assigned to the borehole formation attributes, and the prior probabilities of these cells are set to 1. Then, the constrained Delaunay triangulation method is used to establish the spatial relationships of the boreholes. The borehole wellhead locations are projected onto a two-dimensional plane, and the region is divided based on the attitude simulation results. High-density regions are extracted according to a set threshold, and boreholes within these regions are forcibly connected and prioritized in the triangulation.

[0039] For each simulated grid cell, its prior probability distribution is determined through spatial correlation calculation. The target cell is defined. With drilling The spatial correlation function of the micro-element of stratum k in the middle formation: (1) The influence weight of a single borehole is obtained by integrating the vertical segment of the formation drilled through the borehole: (2) Accumulate the influence weights of all boreholes within the triangular region: (3) Finally, the prior probability of a cell belonging to stratum k is obtained through normalization. The stratum attribute corresponding to the highest prior probability is assigned as the final attribute of the cell, thus completing the prior model construction of the entire grid field.

[0040] (4) Step four, the specific simulation optimization path is as follows: The simulated path optimization is achieved through the following technical process: First, deterministic boreholes are identified based on deterministic profiles generated from attitude simulation. In the triangulation network, triangles are categorized into four types—fully constrained, double-constrained, single-constrained, and unconstrained—based on the number of deterministic vertices they contain. Subsequently, a quantitative evaluation system is established, calculating a comprehensive priority index using equilateral similarity and area standardization indices.

[0041] Let the three interior angles of the triangle be A, B, and C. Then the interior angle deviation is... Calculated using equation (5). Let the lengths of the three sides of the triangle be a, b, and c, and the average side length be... And the deviation of the side length is The ESI is calculated using equations (7) and (8). Considering the equal contribution of interior angles and side lengths to shape regularity, equal-weight coupling is used to obtain the ESI.

[0042] (5) (6) (7) (8) (9) The formula for calculating the area standardization index is: (10) The overall priority index is determined by the following formula: (11) The final simulation path is executed in the following order: first, the triangles are classified by type; second, the triangles of the same type are arranged in descending order of PI value; and finally, within the same triangle, the simulation is performed in descending order of prior probability.

[0043] Step 5: Local scanning strategy based on attitude simulation Local scan parameters are defined using a three-dimensional scaling factor vector combined with the scale of the formation bounding box. For the element to be simulated... Determine its bounding box based on the stratigraphic attribute k. This establishes the local scanning range. The scanning range is the principal direction vector of the bounding box. Length parameter and center coordinates A jointly defined cuboid region. During the scanning process, in addition to the current bounding box... In addition, it is necessary to include n adjacent bounding boxes, where the value of n is determined according to the actual modeling requirements and is associated with the maximum number of similar samples. A standard scanning template is used for pattern search, and the scanning range is gradually expanded from near to far. The scanning process is terminated when the preset maximum number of samples is reached.

[0044] Step six, automatic simulation, specifically involves: Based on the aforementioned technical process, the automatic simulation in this implementation scheme is performed according to the following steps: First, a simulation grid is established and a prior probability model is calculated. Borehole resampling points and other deterministic data are assigned to corresponding grid cells, which, together with the high-density region determined by the attitude simulation, constitute a deterministic constraint dataset. Subsequently, the prior probability distribution of the remaining grid cells is calculated based on spatial correlation, thus completing the construction of the prior model.

[0045] The beneficial effects of this embodiment are as follows: First, principal component analysis is used to extract the strata extension direction from discrete boreholes and establish a modeling space. After discretization using a regular grid, regions are divided along the dominant direction and high-density areas are identified, thereby constructing a simulated exploration profile. After profile filling is completed based on machine learning, a contour reconstruction algorithm is used to generate training images. Subsequently, constrained Delaunay triangulation is used to establish borehole spatial relationships, quantify the stratigraphic correlation between grid cells and boreholes, and construct a simulation path optimization strategy by combining equilateral similarity and area normalization index to determine the simulation priority of regions with different constraint levels. During the simulation execution phase, the local scanning range of the attitude constraint is defined according to the stratigraphic bounding box parameters to ensure that the sampling mode fully reflects the stratigraphic variation characteristics. Finally, all grid cells are simulated sequentially according to the optimized path to generate a three-dimensional model that conforms to geological laws. This method effectively overcomes the dependence of traditional methods on training images through three innovations: automatic simulation of strata attitude, path optimization, and local scanning, and significantly improves the accuracy and reliability of complex stratigraphic modeling under sparse data conditions.

[0046] The following is a specific implementation example of this application: This embodiment proposes a data-driven method for high-precision attitude simulation and 3D geological MPS modeling. This method integrates machine learning profile generation and contour reconstruction techniques to achieve automatic construction of training images and unit-level attitude constraints, providing a solution for accurate modeling of complex geological structures under sparse data conditions.

[0047] This embodiment constructs a high-precision three-dimensional geological model based on borehole data from a mining area, illustrating the implementation process of this technical method and its efficiency and practicality in application. The following description, in conjunction with the accompanying drawings, further explains this embodiment. The method steps and technical route of this application are as follows: Figure 2 As shown.

[0048] (1) Collect wellhead coordinates, eccentricity data, and formation stratification data of all boreholes in the study area as input data, such as Figure 3 As shown, the main extension direction of the rock strata was extracted from the spatial distribution of boreholes using principal component analysis, and this direction was defined as the X-axis of the modeling space. The modeling boundary was determined based on the principles of two-dimensional geostatistics and bounding box theory. The topographic surface of the study area was fitted using borehole head coordinates, and this surface was used to cut the modeling boundary, establishing an initial three-dimensional stratigraphic model containing borehole information. A regular mesh was used to discretize the model, with the mesh cell size set to 10m × 10m × 5m according to the borehole density, ensuring that each mesh cell contains at most one borehole.

[0049] (2) Divide the modeling space into 60 equally spaced sub-regions (X=K, K∈[1,60]) along the X-axis, and count the number of boreholes in each sub-region. Set a borehole count threshold of 5, and identify sub-regions exceeding this threshold as high-density regions. The analysis method is as follows: Figure 4 As shown. In these high-density areas, a support vector machine classifier is used to predict the stratigraphic properties of unfilled grid cells, obtaining complete simulated exploration profile data. These two-dimensional profiles are then transformed into three-dimensional deterministic geological models using a contour reconstruction algorithm. These deterministic models are used as training images for subsequent MPS simulations, and their construction process is as follows. Figure 4 As shown.

[0050] (3) The constrained Delaunay triangulation method is used to establish the spatial relationship of the boreholes, project the boreholes onto a two-dimensional plane and identify high-density regions. The construction process is as follows: Figure 5 As shown. For each simulated grid cell, its prior probability distribution is determined through spatial correlation calculation, as described in the following method. Figure 6 As shown, the influence weight of a single borehole is obtained by integrating the vertical segment of the borehole formation. The influence weights of all boreholes within the triangular region are then summed, and the prior probability distribution of each unit is obtained through normalization. The formation corresponding to the maximum prior attribute of each unit is taken as its prior formation attribute, thus completing the construction of the prior model.

[0051] (4) Based on the triangulation results, triangles are classified into four categories according to the number of deterministic vertices they contain: fully constrained, double-constrained, single-constrained, and unconstrained. For triangles of the same type, a comprehensive priority index PI = α × ESI + (1-α) × (1-ANI) is constructed, where ESI is the equilateral similarity index, ANI is the area standardization index, and the weighting coefficient α is 0.6. The simulation priority is determined by arranging the triangles in descending order of triangle type and PI value to ensure that high deterministic regions are simulated first.

[0052] (5) The occurrence is defined using a series of stratigraphic bounding boxes, such as Figure 7 As shown. The local scan range is defined based on the formation bounding box parameters, and the process is as follows: Figure 8 As shown. For the element to be simulated, the local search area is determined based on the bounding box of its stratum, using the scaling factor vector λ=[0.5,0.3,0.2]. During the scanning process, in addition to the current bounding box, three adjacent bounding boxes are also included to ensure that the sampling mode can fully reflect the stratigraphic variation characteristics.

[0053] (6) After completing the above preparations, perform MPS simulation within the local search range of the attitude constraints according to the optimized simulation path. First, process the deterministic constraint data, and then, based on the prior probability distribution and local scanning strategy, sequentially complete the formation property simulation of all grid cells. Some of the established 3D training images are shown below. Figure 9 As shown, the final generated three-dimensional geological model is as follows: Figure 10 As shown, through comparative analysis with the verification borehole, it is demonstrated that the proposed method can still maintain high modeling accuracy under sparse borehole data conditions.

[0054] Those skilled in the art will readily understand that the above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A three-dimensional geological MPS modeling method based on high-precision attitude simulation, characterized in that, Includes the following steps: S10, Obtain and extract the main extension direction of the rock strata from the borehole data of the study area, establish a modeling space based on the main extension direction of the rock strata, and then discretize the modeling space into grid cells; S20, based on the main extension direction of the rock strata, the modeling space is divided into multiple sub-regions, high-density regions with more than a preset threshold of boreholes are identified, a machine learning classifier is used in the high-density regions to predict the stratigraphic properties of unfilled grid cells, a simulated exploration profile is generated, and a training image is constructed based on the simulated exploration profile. S30, Based on the borehole data, establish the borehole spatial relationship through constrained Delaunay triangulation, and calculate the prior probability distribution of the mesh cells to construct a prior model; S40, based on the prior model and training images, the simulation path is optimized through a quantitative evaluation system, and the local scanning range of the attitude constraint is defined based on the stratigraphic bounding box parameters; S50 performs multi-point geostatistical (MPS) simulations according to the optimized simulation path and the local scan range constrained by attitude, generating a three-dimensional geological model.

2. The three-dimensional geological MPS modeling method based on high-precision attitude simulation as described in claim 1, characterized in that, Step S10 is as follows: The wellhead coordinates, inclination data, and formation stratification data of all boreholes in the study area were obtained as input data. Principal component analysis is used to extract the main extension direction of rock strata from the spatial distribution of boreholes, and this direction is defined as the X-axis of the modeling space. The modeling boundary is determined based on the principles of two-dimensional geostatistics and bounding box theory. The topographic surface of the study area is fitted using the borehole wellhead coordinates. The modeling boundary is then cut through this surface to establish an initial three-dimensional stratigraphic model containing borehole information. The initial three-dimensional formation model is discretized into grid cells using a regular grid. The size of the grid cell is set according to the borehole density, and each grid cell contains at most one borehole.

3. The three-dimensional geological MPS modeling method based on high-precision attitude simulation as described in claim 1, characterized in that, Step S20 is as follows: The modeling space is divided into multiple equally spaced sub-regions along the X-axis, and the number of boreholes in each sub-region is counted. Set a threshold for the number of boreholes, and identify sub-regions exceeding this threshold as high-density regions; In high-density areas, a support vector machine classifier is used to predict the stratigraphic properties of unfilled grid cells to obtain complete simulated exploration profile data. The simulated exploration profile data is transformed into a three-dimensional deterministic geological model using a contour reconstruction algorithm, and this deterministic geological model is used as the training image.

4. The three-dimensional geological MPS modeling method based on high-precision attitude simulation as described in claim 1, characterized in that, Step S30 is as follows: Based on the borehole data, the constrained Delaunay triangulation method is used to establish the spatial relationship of the boreholes, project the boreholes onto a two-dimensional plane and identify high-density areas. For each simulated grid cell, its prior probability distribution is determined by spatial correlation calculation; The influence weight of a single borehole is obtained by integrating the vertical segment of the borehole formation. The influence weights of all boreholes within the triangular region are accumulated, and the prior probability distribution of each unit is obtained through normalization. The formation corresponding to the maximum prior attribute of each unit is taken as its prior formation attribute, thus completing the construction of the prior model.

5. The three-dimensional geological MPS modeling method based on high-precision attitude simulation as described in claim 1, characterized in that, In S40, the optimized simulation path is specifically as follows: Based on the constrained Delaunay triangulation results, triangles are classified into four categories according to the number of deterministic vertices they contain: fully constrained, double-constrained, single-constrained, and unconstrained. For similar triangles, a comprehensive priority index PI is constructed; The simulation priority is determined by sorting the triangle types and PI values ​​in descending order to ensure that high-deterministic regions are simulated first.

6. The three-dimensional geological MPS modeling method based on high-precision attitude simulation as described in claim 5, characterized in that, The formula for calculating the comprehensive priority index PI is as follows: PI = α × ESI + (1 - α) × (1 - ANI) Where ESI is the equilateral similarity index, ANI is the area normalization index, and α is the weighting coefficient; The formula for calculating the Equilateral Similarity Index (ESI) is as follows: ) in, To normalize the interior angle deviation, This refers to the deviation in side length. The formula for calculating the Area Standardized Index (ANI) is as follows: in, This represents the area of ​​the current triangle. The minimum area among all triangles, It represents the maximum area among all triangles.

7. The three-dimensional geological MPS modeling method based on high-precision attitude simulation as described in claim 6, characterized in that, The normalized interior angle deviation The calculation formula is: in, The formula for calculating the interior angle deviation is: Among them, A, B, and C are the three interior angles of the triangle; The side length deviation The calculation formula is: Where a, b, and c are the lengths of the three sides of the triangle; For the average side length, .

8. The three-dimensional geological MPS modeling method based on high-precision attitude simulation as described in claim 1, characterized in that, In step S40, the local scanning range defined by the stratigraphic bounding box parameters for attitude constraints includes: for the unit to be simulated, the local scanning area is determined by the bounding box of its stratum and the scaling factor vector; during the scanning process, in addition to the current bounding box, several adjacent bounding boxes are also included to ensure that the sampling mode can fully reflect the stratigraphic variation characteristics.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by a processor, the computer instructions implement the steps of the method as described in any one of claims 1 to 8.

Images